Properties Of Robust Primary Component Estimation And Utilization In Multi-Fact Pricing Model

As we all know, we make the least square estimator of the parameter of linear model under the condition that the original data must submit to Gauss-markov theory. If the data does not meet the need, least square estimation will result in fake-regression. In the case of this, Principal estimation and M-estimation are adopted to conquer multi-collinearity and outliers. But they two cannot perform well when data is of the two kinds of illness.This paper comes up with a new method named robust principal estimation to estimate parameter of linear model which can deal with multi-collinearity and outliers simultaneously. The paper demonstrates it is superior to the principal estimation and M-estimation. As a example, We finally adopt the new method to estimate parameter of Multivariate Linear Model and make comparison among the three estimations in order to test the demonstration.